Skip to content

Feature Intake & Single Source of Truth

Purpose: Fix the root cause behind the timeline disconnect - new feature ideas (mostly from the Claude/Cloud Design) arriving mid-cycle and being treated as already-specced work. This page defines the one place the plan lives, and the gate every new idea passes through before it becomes development.

The single source of truth

The Feature Matrix is the one source of truth for scope, priority, and status. Not the Cloud Design, not the prototype, not Slack threads, not a fast-moving MkDocs page.

  • The Cloud Design is a visual reference (great for what it should look like), not a spec and not a status tracker.
  • The prototype is the reference for the shipped core, historical.
  • Linear is the execution system; each Feature Matrix row maps to a roadmap-epic in Linear.

If it isn't in the Feature Matrix with an owner, a priority, and a status, it isn't committed work.

The intake gate: capture, review, lock, build

Every new feature request (including Claude-design screens) flows through four steps. Nothing enters a development cycle until it is locked.

  1. Capture. The idea is logged in one place (a Linear issue with the intake label + a link to the Claude-design screen if there is one). Raw capture only - no estimate, no cycle yet.
  2. Review. The team scopes it: what it touches, dependencies, rough effort (S/M/L/XL), and whether it splits into slices. Divergences from the shipped build are flagged. Output = a scoped issue with acceptance criteria.
  3. Lock. Conor signs off on scope + priority + which wave. Once locked, it does not change mid-cycle. New thoughts go back to step 1 for the next planning boundary - they do not reopen locked work.
  4. Build. Only locked items enter a development cycle. A feature is done when it clears QA (multi-pass), not when the first PR merges.

Why this exists

  • Mid-cycle scope changes destabilize QA: a feature under test changes, has to be re-tested, and the cycle slips. A locked scope + a review boundary is what lets us give reliable dates.
  • A major feature now spans multiple cycles (4+ weeks) with a QA loop. That is the cost of stability, and it only works if scope holds still while it's being built and tested.
  • One source of truth means Conor, the team, and Linear never disagree about what's committed.

Design sync + page sign-off

The root cause of the rework Conor has seen: Cloud Design pages get implemented mid-cycle with no formal sign-off, so the dev builds against a moving target and the client's intent is guessed at.

  • Bi-weekly design sync with Conor (designer) + the dev who will build the page + the product lead (Ivan S). Review the finalized Cloud Design pages, ask domain questions, and freeze/sign off each page before it enters build. A signed-off page does not change until the next planning boundary.
  • Cloud Design to Linear intake: when a page is finalized, push it to the Linear intake queue with a screenshot + notes; the assigned dev participates in sign-off and asks clarifying questions there.
  • The first design sync is a working takeoff-domain Q&A - the dev who owns takeoff asks Conor directly. This models the question-asking Conor wants and closes the "devs don't understand the domain" complaint head-on.
  • Conor gets access to the intake queue only, not the whole project.

Cadence

  • Standing weekly syncs are replaced by topical deep-dives and the bi-weekly design sync above - scheduled per subject, with the right owner in the room and a pre-circulated agenda, only when a decision needs it.
  • Planning boundaries (cycle starts) are where new locked work enters. Between boundaries, the cycle scope is frozen.
  • Single source of truth = the adopted roadmap: the Module Roadmap + Master Feature List CSV, mirrored by the Feature Matrix. The Cloud Design stays the picture of how it should look; it is not the spec or the status. (The live plan is now the Project Timeline (Gantt) + 8-Week Lock Status.)

Proposal status

This process is a proposal for Conor (see the Conor Briefing). Once he agrees, it becomes how we run intake, and the Feature Matrix becomes the artifact he and the team burn through together.